OCR command-line tool

Starting in 11.1, Foxit PDF SDK provides a command-line OCR tool on Windows and Linux. Use it to process single pages, whole PDFs, or scanned PDFs to editable formats without integrating the OCR API directly.

For OCR API capabilities, see OCR (Optical Character Recognition). For OCR resource packages and directory layout, see OCR sample setup and run guide.

Use cases

The OCR command-line tool suits:

• Quickly verifying OCR runtime, license, and resource directory setup
• Running OCR from scripts, schedulers, or batch pipelines
• Invoking from Java, Python, C#, Node.js, or other hosts via a separate process
• Outputting scanned PDFs directly as editable formats

How to obtain

Obtain the tool the same way as OCR resources:

• Request a trial or download instructions from the Foxit Developer Hub
• Contact Foxit technical support or sales for the OCR command-line tool package

Prerequisites

Before use, confirm:

• Foxit PDF SDK runtime libraries are deployed correctly
• License files are ready: gsdk_sn.txt, gsdk_key.txt
• OCR engine resource directory is prepared
• The tool executable, SDK dependencies, and license files are in the same working directory

Prepare the run directory

Create a dedicated directory, for example OCRExeTool.

Windows

Place these files in the same directory:

• ocr_win64.exe
• fsdk_win64.dll
• gsdk_sn.txt
• gsdk_key.txt

Linux x64

Place these files in the same directory:

• ocr_linux64
• libfsdk_linux64.so
• gsdk_sn.txt
• gsdk_key.txt

Then view help:

ocr_win64.exe --help

./ocr_linux64 --help

Command format

Windows:

ocr_win64.exe [options]

Linux x64:

./ocr_linux64 [options]

Required parameters

Parameter	Description
-type	OCR mode: 0 = OCRPDFPage, 1 = OCRPDFDocument, 2 = OCRConvertTo
-input	Input PDF path
-output	Output file path
-engine	OCR engine resource directory path

Common optional parameters

Parameter	Description
-lang	OCR languages, e.g. "English, Chinese-Simplified"
-edit	Whether output is editable: yes / no
-pw	Password for the input PDF
-range	Page range, e.g. 0,2
-format	Valid only when -type 2. Output format: 0=DOCX, 1=DOC, 2=RTF, 3=XLSX, 4=XLS, 5=PPTX, 6=HTML

-type maps to OCR APIs as follows:

• 0: OCRPDFPage
• 1: OCRPDFDocument
• 2: OCRConvertTo

Image and recognition parameters

These map mainly to OCRConfig:

Parameter	Description
-is_detect_pictures	Detect pictures: yes / no
-is_remove_noise	Denoise: yes / no
-is_correct_skew	Deskew: yes / no
-is_enable_text_extraction_mode	Enable text extraction mode
-is_sequentially_process	Process pages sequentially
-is_auto_overwrite_resolution	Auto-overwrite resolution
-resolution_to_overwrite	Manual resolution, e.g. 300
-confidence	Confidence threshold, 0-100
-ignore_image_width	Minimum width to ignore small images
-ignore_image_height	Minimum height to ignore small images

Parameter tips

• Single page: -type 0
• Whole PDF: -type 1
• OCR then Word, Excel, PowerPoint, or HTML: -type 2 with -format
• Editable output: -edit yes
• Limit pages: -range
• Noisy pages: -is_remove_noise yes
• Skewed pages: -is_correct_skew yes
• Wrong resolution metadata: disable -is_auto_overwrite_resolution and set -resolution_to_overwrite
• Filter low-confidence results: increase -confidence

Examples

Windows

ocr_win64.exe -type 0 -input E:/TestFiles/AboutFoxit_ocr.pdf -output E:/TestResult/OCRPDFPage_AboutFoxit_ocr_edit_yes.pdf -engine E:/OCREngine/x64 -lang English -edit yes -is_detect_pictures yes -is_remove_noise yes -is_correct_skew yes -is_enable_text_extraction_mode yes -is_sequentially_process yes -is_auto_overwrite_resolution yes -resolution_to_overwrite 300 -confidence 38 -range 0,2 -ignore_image_width 20 -ignore_image_height 20

Linux x64

./ocr_linux64 -type 0 -input /home/test/AboutFoxit_ocr.pdf -output /home/test/result/OCRPDFPage_AboutFoxit_ocr_edit_yes.pdf -engine /opt/OCREngine/x64 -lang English -edit yes -is_detect_pictures yes -is_remove_noise yes -is_correct_skew yes -is_enable_text_extraction_mode yes -is_sequentially_process yes -is_auto_overwrite_resolution yes -resolution_to_overwrite 300 -confidence 38 -range 0,2 -ignore_image_width 20 -ignore_image_height 20

Calling from Java

If you do not embed the OCR API, invoke the tool via an external process:

import java.io.File;
import java.io.IOException;

public class RunOCRTool {
    public static void main(String[] args) {
        try {
            ProcessBuilder processBuilder = new ProcessBuilder(
                "E:\\OCRExeTool\\ocr_win64.exe",
                "-type", "0",
                "-input", "E:\\TestFiles\\AboutFoxit_ocr.pdf",
                "-output", "E:\\TestResult\\OCRPDFPage_AboutFoxit_ocr_edit_yes.pdf",
                "-engine", "E:\\OCREngine\\x64",
                "-lang", "English",
                "-edit", "yes"
            );

            processBuilder.directory(new File("E:\\OCRExeTool"));
            processBuilder.redirectOutput(ProcessBuilder.Redirect.INHERIT);
            processBuilder.redirectError(ProcessBuilder.Redirect.INHERIT);

            Process process = processBuilder.start();
            int exitCode = process.waitFor();
            System.out.println("Program execution completed, exit code: " + exitCode);
        } catch (IOException | InterruptedException e) {
            e.printStackTrace();
        }
    }
}

Batch processing tips

• Validate a single file end-to-end before scaling to batch jobs
• Split work by file and let an external scheduler handle concurrency
• Tune concurrency for CPU, memory, and license limits
• Collect stdout, stderr, and exit codes for failed tasks

For in-app batch processing across documents or page ranges, consider OCRPDFDocuments in the OCR API.

Exit codes

Code	Description
0	Success
100	Unknown error
101	Invalid license
102	Invalid parameters
103	GSDK initialization failed
104	Invalid OCR engine directory
105	OCR engine initialization failed
106	Document load failed

Command-line tool vs OCR API

• Quick runs, scripts, external processes: OCR command-line tool
• Embedded logic, fine-grained callbacks, direct SDK objects: OCR (Optical Character Recognition)

To verify resources, license, and the basic pipeline, run the tool or sample first, then decide whether to integrate the API.